基于自适应评价的非线性系统神经网络控制  

Neural network control of nonlinear systems based on adaptive critic

作　　者：陈宗海[1] 文锋[1] 王智灵[1] 

机构地区：[1]中国科学技术大学自动化系,合肥230027

出　　处：《控制与决策》2007年第7期765-768,773,共5页Control and Decision

基　　金：国家自然科学基金项目(60575033).

摘　　要：针对一类非线性系统,提出了一种自适应评价方法.该方法可以控制系统输出对参考信号进行跟踪,其评价函数可直接解析求出.该方法只需一个动作网络用于产生控制动作,并且方法中的网络权值初始化可随机选取.使用Lyapunov方法对整个系统的动态性能进行分析,证明了在一定条件下此方法能保证闭环误差及网络权值一致最终有界.仿真结果与理论分析相一致,证明了所提出方法的有效性.A novel adaptive critic method is proposed for a class of nonlinear systems, by using which system output is controlled to track a reference trajectory. In this method, critic function can he solved analytically and only one action network is needed to generate control action. Moreover, network weights are initialized randomly. The analysis of dynamic performance of the overall system is performed by using Lyapunov method, which proved that by using the proposed method uniform ultimate boundness of close-loop error is guaranteed under certain conditions. Simulation results are consistent with theoretical analysis and show the effectiveness of the method.

关 键 词：自适应评价 神经网络 LYAPUNOV方法 强化学习 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]